The Agent Skills Directory

[COMMAND_EXECUTION]: The script scripts/agent.py performs administrative system modifications using iptables for network containment and powershell for identity management, which can lead to significant infrastructure impact if misused.\n- [REMOTE_CODE_EXECUTION]: The functions disable_ad_account and reset_ad_password in scripts/agent.py are vulnerable to command injection. They construct PowerShell command strings using f-strings with unsanitized user-provided input, which is then executed via powershell -Command. This allows an attacker to execute arbitrary PowerShell commands by providing a crafted username.\n- [PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection. Ingestion points: Untrusted incident telemetry and scope data (IPs, account names) processed in the incident response workflow. Boundary markers: None identified. Capability inventory: High-impact administrative control via iptables and Active Directory management. Sanitization: No input validation or sanitization is performed on external data before it is passed to system commands.

containing-active-breach